After a node becomes 'online' from a period of downtime, it looks like the Availability statistic isn't updated properly. The node is inappropriately at 0% availabiality, when there is 0% ping loss just as soon as the node comes up.
This is -really- hard to explain, so I'll just add a screenshot.
I'm writing a query that excludes change management timeframes from a critical device availability statistic, and this is causing a 0.03% margin of error
